The biggest challenge in hardware design will be to cut system power dissipation. For example, a small, 4x4 antenna array with 16 elements is likely to consume 3 to 4W, using today’s best case estimates. Special designs, such as sparse arrays in which the array elements are more widely spaced than in regular arrays, will make it possible to build such antennas and cool them passively.
But we need to do more. We can increase the efficiency of the power amplifiers by using new techniques that do not require power-hungry linearization. We can reduce the power consumption of data converters by reducing their bit depth and making use of RF DACs. Researchers should also work on validating the assumption that digitization noise, phase noise and spurious signals from the amplifiers average out in the far field, so we can design 5G equipment with simpler, more energy-efficient components.
Another large R&D challenge is the assembly of mm-wave antenna array panels. We need new techniques that allow sufficient cooling, guarantee the required system performance and are suitable for low-cost and high-volume manufacturing. The key issue will be to connect the semiconductor chips electrically to the antennas as well as mechanically to the cooling plate - existing solutions are too expensive for use in a 5G roll-out.
There’s work to be done on the antennas themselves, too, especially the influence of mutual coupling between antennas on the performance of the power amplifiers. This coupling shifts the load impedance, causing an unwanted phase shift in the power amplifier. This distorts the beam it is helping to form and may also affect the reliability of the semiconductor devices in it. So we need to develop inter-antenna isolation and/or correction algorithms, plus load-insensitive power amplifier architectures to solve the problem.
Low noise amplifier designs need updating, too. The LNAs in an array antenna receive all signals from all directions and so can distort any particularly strong signals they receive, desensitizing the receiver. Therefore we need LNAs with high linearity and the lowest noise figures possible –which is difficult to achieve at low power consumption as well.
We’re still awaiting 5G standardization, but while that happens we can be working on better signal filtering approaches, even though the requirements may not be as severe as in current wireless infrastructure systems. Integration of filters, either on-chip or in the antenna panel board, will be needed to meet size constraints. We also need to develop new filtering techniques at mm-wave frequencies that are small and can be integrated at low cost.
Finally, we need to do more work on antenna array architectures that can steer multiple beams simultaneously to several users at once. Such panels make use of spatial diversity, which puts requirements on the beam shape and the beam steering. When antenna arrays get big, to allow them to be passively cooled, beams get small – so the requirements throughout the system to enable accurate beam steering may be particularly hard to meet.
Which of these challenges is most important? My money is on two: fixing the power consumption issue,